No abstract

No abstract

The present work deals with the numerical prediction of the post buckling progressive and final failure response of stiffened composite panels based on structural nonlinear finite element methods. For this purpose, a progressive failure model (PFM) is developed and applied to predict the behaviour of an experimentally tested blade-stiffened panel found in the literature. Failure initiation and propagation is calculated, owing to the accumulation of the intralaminar failure modes induced in fibre reinforced composite materials. Hashin failure criteria have been employed in order to address the fiber and matrix failure modes in compression and tension. On the other hand, the Tsai-Wu failure criterion has been utilized for addressing shear failure. Failure detection is followed with the introduction of corresponding material degradation rules depending on the individual failure mechanisms. Failure initiation and failure propagation as well as the post buckling ultimate attained load have been numerically evaluated. Final failure behaviour of the simulated stiffened panel is due to sudden global failure, as concluded from comparisons between numerical and experimental results being in good agreement.

Condition monitoring (CM) of ship hull structures is a promising field that has recently attracted the interested of researches. The main challenge behind CM is to develop a system that gets as input sensor readings from the structure and provide the damage locus as an output. In this regard, the current study proposes two alternative CM digital twin schemes for solving this inverse engineering problem. The first one is based on a Finite Element (FE) – Optimization cooperative framework that solves several times the model until the predicted strains match the measured ones and as such the damage location has been found. The other scheme is based on a cooperative framework of Artificial Neural Networks (ANNs) used for classification and fitting, that may be regarded as surrogated models which provide solutions instantaneously. The ANNs are trained through the numerical solutions provided by the FE model. A thin-walled hollow cantilever beam, that resembles a hull-girder subjected to principal stresses under vertical bending, has been adopted. The performed work has allowed for the selection and evaluation of the locations for sensor placement and the estimation of the damage sensitive area for monitoring. Both CM digital twin schemes have proven to be promising for the theoretical simplified examined case.

No abstract